Impact of thermal desorption tubes on the variability of exhaled breath data

Due to the overall low abundance of volatile compounds in exhaled breath, it is necessary to preconcentrate the sample prior to traditional thermal desorption (TD) gas chromatography mass spectrometry analysis. While certain aspects of TD tubes, such as volatile storage, have been evaluated, many aspects remain uncharacterized. Two common TD tubes, Tenax TA and Biomonitoring 5TD tubes, were evaluated for background content and flow rate variability. The data illustrate that the Biomonitoring 5TD tubes have the highest number (23) and abundance of background contamination greater than 3x the mean noise when compared to Tenax TA (13) and empty tubes (9). Tentative identifications of the compounds in the background contamination experiment show that greater than 59% (16/27) of the compounds identified have been reported in the breath literature. The data illustrate the TD tube background abundance could account for more than 70% of the chromatographic signal from exhaled breath for these select compounds. Flow rate measurements of 200 Tenax TA and 200 Biomonitoring 5TD tubes show a large range in measured flow rates among the TD tubes (Tenax: 252.9–284.0 ml min−1, 5TD: 220.6–255.1 ml min−1). Finally, TD tubes of each type, Tenax TA and Biomonitoring 5TD, previously established to have high, medium, and low flow rates, show insignificant differences (p > 0.05) among the tubes of different flow rates, using both gas standards and an exhaled breath from a peppermint experiment. Collectively, these results establish overall background compounds attributed to each TD tube type tested. Additionally, while measured flow rate variability is present and plausibly impacts exhaled breath results, the data demonstrate no statistically significant difference was observed between tubes showing high, medium, and low flow rates from two separate sample types.


Introduction
In recent years, analysis of trace amounts of volatile compounds in exhaled breath for the non-invasive detection of disease, such as asthma, and physiological processes, like hypoxia, is a growing field of research [1][2][3][4][5].While Pauling noted more than 200 compounds in human breath in the 1970s, exhaled breath analysis has ultimately shown limited clinical transition [6].It is hypothesized that the lack of exhaled breath clinical assays is a result of an absence of standardization in sampling and analysis.
Although exhaled breath is on-demand and readily available, it has been shown that the sampling approach can have a significant impact on the results [7][8][9][10][11][12][13]. For example, Sukul et al demonstrated an effect on exhaled breath volatiles in relation to breath holding and ergonomic position while sampling [9,11].Furthermore, it has been suggested that controlled sampling for end-tidal breath provides more concentrated and less variable samples than a mixed expiratory breath due to a lack of dead space dilution [14].As exhaled breath contains low concentrations of analytes, it is necessary to consider the sample collection methodology to reduce external sources of variability from the analysis and improve reporting of identified volatiles.
One of the most common methods of exhaled breath sample collection for volatile analysis is by thermal desorption (TD) tubes.Sampling onto TD tubes can be performed passively or actively.For passive collection, the TD tube is left uncapped in the sampling environment for a proscribed period of time while volatile compounds diffuse onto the sorbent material in the tube.Active collection occurs when a pump is connected to the TD tube and the sample, such as ambient air or gas from an exhaled breath bag, is pulled across the adsorbent material.TD tubes utilized for volatile compound preconcentration have several advantages.First, depending on the specific adsorbent(s), TD tubes can frequently collect a wide range of volatile organic compounds (VOCs, C 2 -C 40 ) allowing for detection levels in the ppb/ppt range on most gas chromatography mass spectrometry (GC-MS) instruments.Second, the TD tubes and adsorbents are simple to use, easily cleaned, and reusable.Finally, TD tubes allow for off-site sampling in relation to the analytical lab, and ultimately storage, due to the prolonged stability of volatiles adsorbed within the tubes [15,16].
Two types of TD tubes have become popular within exhaled breath research, Tenax TA and the dual bed, Biomonitoring 5TD tubes.Tenax TA, a porous hydrophobic polymer, is ideally suited for retention of lower volatility/higher molecular weight compounds of approximately C 6 -C 30 .As water can negatively affect GC-MS performance, the low affinity of Tenax TA for water makes it an excellent adsorbent for use with high humidity breath samples.Dual bed sorbent tubes, such as the Biomonitoring 5TD tubes, have an extended volatile adsorption range using a combination of adsorbents.Biomonitoring 5TD tubes are comprised of both Tenax TA and carbograph adsorbents which extends the adsorption range to approximately C 3 -C 30, allowing researchers to collect lower molecular weight/higher volatility compounds more reliably from breath at the expense of potential issues with humidity trapping [13].
As the use of sorbent tubes continues to be a primary method for exhaled breath sampling, it is vital to evaluate the impact TD tubes have on the analytical results.Currently, it is assumed that all TD tubes are uniform within a given type of adsorbent regarding manufacturing and performance.However, it is hypothesized that variation among TD tubes, within a specific type, could impact exhaled breath results.In this study, the effect of trace background compounds and other sources of TD tube variability are evaluated for their influence on standard and exhaled breath data by monitoring both endogenous and exogenous VOCs.

TD tubes
Four types of TD tubes, empty inert, empty stainless steel (SS), SS Tenax TA 35/65 mesh (Tenax); and stainless-steel Biomonitoring TD5149 (5TD), were purchased from Markes International (South Wales, UK).The empty inert and empty SS TD tubes contain no adsorbent, with the inert coated with nonreactive silicon.All TD tubes, regardless of adsorbent or absence of adsorbent, were reconditioned prior to and after each use at 320 • C for 2 h with 50 ml min −1 of 99.999% N 2 backflush on a Markes TC-20, as recommended by the manufacturer.
All TD tube masses were determined using a Mettler Toledo AB135-S balance (Columbus, OH, USA).All TD tubes flow rates were measured with a MultiRAE lite pump (calibrated daily for flow and CO 2 , Honeywell, San Jose, CA, USA) and a Mesa Laboratories 510 Defender with the DryCal Pre V Beta 3 software (v.1.1,Lakewood, CO, USA).Please see supplemental data 1 for an image of the flow rate measurement setup.The mean flow rate, determined over approximately 2 min of continuous collection (approximately 30 measurements), was recorded for further analysis.

TD tube background analysis
Approximately 200 of each adsorbent containing tube, Tenax (218) and Biomonitoring 5TD (210), and 50 of each type of empty tube, inert and stainlesssteel, were run on a TD gas chromatography mass spectrometry (TD-GC-MS) system as described in the next section.All tubes were run twice, both with and without TO-14 A internal standard/tuning four component mix (1 ppm each, Restek, Bellefonte, PA, USA).Data files were analyzed for feature alignment and extraction by the Metabolite Differentiation and Discovery Lab (MeDDL) software platform utilizing the settings provided in supplemental data 2 [17,18].To determine the minimum signal settings for the MeDDL software, a section of chromatograms (9.5-9.7 min RT) from 20 representative TD tubes of each type, Tenax TA and Biomonitoring 5TD, empty SS, and empty inert, that was devoid of signal was utilized to determine the noise.The average value from each tube type, was multiplied by 3 (3x) and used for feature selection.Features were tentatively identified via the Thermo Scientific Qual Browser software (v.4.3.73.11) equipped with the NIST 11 library (v.2.4, Gaithersburg, MD, USA, supplemental data 3).
Evaluation of the background contributed from GC and the TD-100xr cold trap were independently performed using the Markes TD-100xr preloaded methods in conjunction with the GC-MS settings in the next section.The ALTEF bag background was assessed by filling five 2 l bags with 99.999% N 2 and sampling 550 ml from each bag onto each TD tube type, Tenax TA and Biomonitoring 5TD (n = 5 each).All bag background samples were run as described in the next section.All data were visualized and evaluated using the Thermo Scientific Qual Browser software and the NIST 11 library.

TD-GC-MS)
All TD-GC-MS analyses were performed in a random order using Microsoft Excel's RANDBETWEEN function (Microsoft, Redmond, WA, USA).Each tube was run on a Markes International TD100xr inline with a Trace Ultra GC connected to an ISQ single quadrupole mass spectrometer (ThermoFisher, Waltham, MA, USA).For injections with internal standards, 1 ml of standard was automatically loaded onto the TD tube, prior to TD, by the TD-100xr.Volatiles were desorbed from the TD tube for 10 min at 310 • C with a flow path temperature of 160 • C. Trapped volatiles were purged with N 2 for 1 min at 50 ml min −1 followed by a secondary desorption at 315 • C for 10 min at 40 • C s −1 with a 5 ml min −1 split flow for an output split of 3.5:1.Volatiles were separated on a Restek Rxi-624Sil MS 60 m × 0.32 mm, 1.8 µm(df) column with a constant flow of 2.0 ml min −1 99.9999% scrubbed helium (Restek, ThermoFisher).The GC temperature was held at 40 • C for 1 min, then ramped to 240 • C at 10 • C min −1 .The temperature was held for 20 min.The ISQ transfer line was operated at a temperature of 230 • C with the ion source at 275 • C. Spectra were acquired over a range of 35-350 amu every 0.15 s.The instrument performance was evaluated, prior to each analytical run, using the EPA TO-15/17 calibrant (27 ppb) compared to six-point calibration curves, as outlined in the EPA TO-17 method [19,20].

TD tube loading from gas sampling bags
Two high concentration 40 l ALTEF gas sampling bags (Restek) were filled with a 1:19 dilution of the EPA TO-15 65 component mix.TO-15 mix was diluted by adding 99.999% N 2 utilizing AliCat Scientific mass flow controllers (Tucson, AZ, USA) into new bags.Each high concentration bag yields 50 ppbv concentration.Additionally, two low concentration 40 l ALTEF bags were prepared by diluting the TO-15 mix with N 2 at a 7:995 (TO-15:N 2 ) ratio to create a concentration of 7 ppbv within the bag.
All TD tubes were loaded with 550 ml of volatiles from the bag, less than 1 h following filling, through direct pumping with a calibrated (daily) MultiRAE [21][22][23].Forty five tubes of each type, Tenax and Biomonitoring 5TD, determined to have high (15 tubes), medium (15 tubes), and low (15 tubes) measured flow rates were loaded, in a random order, from each bag type (high concentration and low concentration).All TD tubes were run and reconditioned, as described above, where reconditioning occurred between each bag type.All data were evaluated for peak areas as described in the previous section.

Participants & exhaled breath collection
The study was approved by the 711th Human Performance Wing Institutional Review Board (FWR20180006H) in accordance with the principles embodied in the Declaration of Helsinki and local statutory requirements.The board's determination of this research as non-human use requires individuals to be informed of the protocol prior to participating, and the ability to quit at any time, but written consent was not required.No metadata was collected from volunteers.Participants were a mix of male and females (>18 years old) who had fasted for at least 1 h prior to sampling.All current local precautions concerning COVID-19 were followed, as defined by the United States Air Force, which could include face masks, social distancing, and other safety measures as required at the time of sampling [24].
Tenax and Biomonitoring 5TD, previously determined to have high, medium, and low measured flow rates, were reconditioned and used for the collection of exhaled breath samples.Lower airway breath was collected using our established protocol, into 5 l ALTEF gas samplings bags from ten subjects after sitting in a relaxed position for at least 5 min (Restek) [2,3,11,25,26].The subjects were then given a Wintergreen Altoid ® , instructed to place it on the tongue, and allow it to dissolve for five minutes (Mars Wrigley Confectionery, Hackettstown, NJ, USA).The mint completely dissolved prior to the end of the five minutes for all subjects.Following, the subjects provided a second 5 l ALTEF bag sample with lower airway breath.A volume of 550 ml was loaded on Tenax and Biomonitoring 5TD, with high, medium, and low flows, from the bags as described above.All volatiles were transferred from the bags on to TD tubes, in a random order, within 10 min of exhaled breath sampling [21][22][23].The tubes were immediately run by TD-GC-MS, as described in a prior section [15].The data were evaluated for compounds relating to the mint and exhaled breath VOCs via the Tracefinder software suite and global VOC analysis by the MeDDL software as described above.

Statistical analysis
Venn diagrams were generated using the Venny online tool [27].Significance for comparing flow rates (high, medium, and low) for each of the four compounds (isoprene, acetone, 1-methyl-4-(1-methylethyl) cyclohexanol (MMECH), and methyl salicylate) on Tenax and Biomonitoring 5TD tubes were calculated using one-way ANOVA in RStudio (v.3.6.3).Principal component analysis (PCA) plots were generated through ggbiplot package within RStudio with ellipses plotting 95% confidence intervals for the three groups.Unpaired one-sample t-test were performed within RStudio and p-values were reported as statistically significant when p < 0.05.All other statistics and figures were made and calculated in GraphPad Prism (v.9.5.1,San Diego, CA, USA).

TD tube background contamination
Although blank or background analyses are frequently run for individual studies, often thorough evaluation of the blank data is over looked.Therefore, to determine the background associated with two of the most frequently used TD tubes for exhaled breath research, Tenax TA and Biomonitoring 5TD TD tubes, approximately 200 of each tube type, without internal standard, were analyzed via TD-GC-MS.Figure 1(A) shows representative scaled total ion chromatograms of empty SS, Tenax SS, and Biomonitoring 5TD SS tubes.The data illustrate a potentially large number of trace levels of compounds are present among TD tubes even those without adsorbent.To further investigate the background compounds attributed to each tube type, a global alignment and feature extraction was applied to the data with settings at three times (3x) the mean noise.Please refer to supplemental data 3 and 4 for all abundances and an overall summary.The Biomonitoring 5TD tubes have the largest number of compounds (23) with abundances above the mean noise threshold (3x) with Tenax TA having 13 and the two empty tubes having eight (inert) and four (SS) respectively (supplemental data 3 and 4).Using the tentative identifications, a Venn diagram was generated (figure 1(B)).The diagram illustrates only two features, tentatively identified as acetaldehyde and acetone, are found in all tube types tested, with or without adsorbent.Furthermore, the data suggest the Biomonitoring 5TD tubes have nine (9) unique features and seven (7) features shared only with tubes containing adsorbent (figure 1(B)).The data in figures 1(A), (B) and supplemental data 3, 4 provide evidence for common background contaminants for both Tenax TA and Biomonitoring 5TD tubes.Furthermore, the results indicate that empty stainless-steel tubes have fewer background features when compared to empty inert coated tubes.
Tenax TA, Tenax GR, and Biomonitoring 5TD tubes are some of the most frequently utilized adsorbents for exhaled breath analysis.Tenax TA, a porous hydrophobic polymer is ideally suited for sampling medium to high boiling point compounds of approximately C 6 -C 30 .It is often selected for exhaled breath analysis due to its hydrophobic nature, i.e., resistant to exhaled breath humidity, and, as illustrated in figure 1, has a relatively low background.However due to the binding range of Tenax, small molecular weight high volatility compounds often remain unbound to Tenax and ultimately are missed in the analysis.As a result, the dual bed Biomonitoring 5TD tubes, comprised of Tenax TA and carbograph adsorbents, have become popular for exhaled breath research due to carbograph's ability to adsorb low boiling point compounds often missed by Tenax.The combination of the two types of adsorbent material makes the Biomonitoring 5TD tubes well suited for the capture of compounds with molecular weights between C 3 and C 30 often found in exhaled breath.However, as illustrated by Wilkinson et al, the Biomonitoring 5TD tubes can be impacted by water from the humidity in exhaled breath [13].Additionally, due to the increased adsorbent range, the Biomonitoring 5TD tubes show an increased number and abundance of background compounds when compared to Tenax TA tubes.While better suited for collection of a broad range of exhaled volatile compounds, the larger background associated with the Biomonitoring 5TD tubes could possibly impact the reported results.
As Tenax and Biomonitoring 5TD have overlapping adsorbent material, it is reasonable that many (7) of their artifacts are shared.Additionally, as both were made of SS, the data show that the compounds attributed to the stainless-steel casing are represented in both TD Tube types (figure 1(B) and supplemental data 3, 4).The inert TD tube casing material shows an additional four background artifacts, compared to the stainless-steel material, which could be attributed to the different TD tube coatings.While the data indicate that the empty tubes contribute to the background in the GC-MS analysis, it is possible that some of the observations made could be attributed to the TD-GC-MS instrumentation, such as the ALTEF bags, TD module, cold trap, or the analytical GC column.To verify that the compounds found in the background data were a result of the TD tubes and not inherent in the setup or TD-GC-MS system, experiments were performed to evaluate the contribution of the GC only, TD trap i.e. no TD tube in place, and the ALTEF bags (supplement data 5).The data indicate little background is attributed to the GC and TD trap compared to blank TD tube samples (supplemental data 5 top).However, the data in supplemental data 5 bottom indicate ALTEF bags may contribute a small number of features not found in blank tube analyses, tentatively identified as acetonitrile, hexamethyl disiloxane, and 2,2-dimethyl decane.While observed, these features do not align with previous reports of ALTEF bag background and additional experimentation is required to thoroughly investigate and corroborate the results [26].Collectively, these data indicate the majority of background features are likely a result of the TD tubes and not from other parts of the TD-GC-MS system or sampling approach.However, as it is likely the results could be different for each analytical system, it is valuable to perform similar procedures to evaluate potential background on discovery instrumentation.While blank injections are often run during analyses to ensure a lack of carryover in the analytical system, addressing compounds found in the blank injections, i.e., subtraction or removal, has not yet been standardized for exhaled breath.Based on the tentative identifications provided in supplemental data 3 and 4 representing blank Tenax TA and Biomonitoring 5TD tubes, greater than 59% (16/27) of the compounds have been reported in the breath literature indicating the compounds were considered as potential targets in the studies [27,28].While it is possible for these compounds to be greater than the background in exhaled breath, such as is often the case with acetone, most breath VOCs are in trace amounts and background could significantly impact the results and reporting.To evaluate the potential contribution of TD tube background compounds on exhaled breath results, abundances from select compounds, found in both the exhaled breath literature and TD tube background evaluation, were tabulated from each TD tube background injections and an exhaled breath mint experiment (discussed further in a future section, supplemental data 3).Figure 2(A) shows the abundances from all of the overlapping compounds from Tenax TA blank tubes and exhaled breath are significantly different (p < 0.05).These data suggest exhaled breath has higher abundances than background although a large portion of the abundances, up to 73%, could be related to the TD tube background.Figure 2(B) illustrates a similar result with the Biomonitoring 5TD tubes where six of the nine compounds show a significant difference from breath to background (p < 0.05) with a possible background contribution of up to 72%.However, three compounds, tentatively identified as acetaldehyde, acetic acid, and toluene show an insignificant (p > 0.05) difference between the Biomonitoring 5TD background and exhaled breath suggesting the abundance observed in exhaled breath could be attributed to the background from the tube.These data suggest the abundance from the background of TD tubes, if selected as a sampling approach, could possibly impact the resulting exhaled breath data.
Overall, the results in figures 1, 2 and supplemental data 3-5 illustrate that there are volatiles likely stemming from the TD tube adsorbent or casing materials which could possibly impacting exhaled breath TD-GC-MS analyses.Several of these VOCs associated with the background have been frequently identified in exhaled breath literature [29][30][31].While no information on sampling procedures was provided in the literature reviews, it is possible background from TD tubes could have contributed to compound reporting [29][30][31].Therefore, background data, generated from blank TD tubes, should be considered in the analysis and reporting of compounds from exhaled breath TD-GC-MS analyses.As these results could be different for each analytical system, it would be recommended to perform similar procedures to evaluate potential background on discovery instrumentation.Regardless, collectively the data indicate that consideration of blank injections and background samples derived from the sampling adsorbent should be appropriately handled in exhaled breath compound reporting.Additionally, it would be beneficial for the exhaled breath community to collectively establish guidelines for handling of blank/background data within volatile biomarker discovery efforts.

Effects of flow rate
As shown with potential for background contamination (figures 1, 2 and supplemental data 3-5) arising from TD tubes, further investigation into potential sources of data variability, such as flow rate, mass, age, and reuse, resulting from TD tubebased sampling requires evaluation.To begin, more than 200 of each TD tube type, Tenax (205) and Biomonitoring 5TD (208), were evaluated for flow rate as depicted in supplemental data 1.The data show the range of the flow rates among both tube types is 252.9-284.0ml min −1 for Tenax TA and 220.6-255.1 ml min −1 for Biomonitoring 5TD tubes (figure 3(A)).Due to the wide range in flow rates observed, three factors were hypothesized to impact the measured flow rates, TD tube mass, the number of reconditioning cycles, and/or age of the tubes.
First, the mass of each TD tube was measured and plotted based on age of the tube.The data in figure 3(B) illustrate Tenax tubes have lower mass variability when compared to the Biomonitoring 5TD tubes.However, a linear fit of the mass by mean flow rate for each tube type shows an insignificant linear correlation between the two measurements (Tenax: p = 0.2034, 5TD: p = 0.1224, figure 3(C)).To determine if the mass variation observed was plausibly a result of differences due to the TD tube casings, 50 empty stainless-steel and 50 empty inert tubes were weighed (supplemental data 6(A)).The data suggest the differences in the masses of the empty tubes are not statistically significant (p = 0.1029) and likely does not explain the differences in the variability in the tube masses observed in figure 3(B).Next, it was hypothesized that the number of desorption/reconditioning cycles could affect the performance of TD tubes due to adsorbent degradation.Utilizing our in-house TD tube tracking system, the number of reconditioning cycles was plotted by tube age (supplemental data 6(B)).The data show, for the 200 tubes of each type evaluated, a downward trend is observed, i.e. with age comes a higher number of uses, with no tube having more than 50 reconditioning cycles.Furthermore, when the number of reconditioning cycles was plotted against mean flow rate, the data illustrate an insignificant linear relationship between the two variables (Tenax: p = 0.1359, 5TD: p = 0.8417, figure 3(D)).Finally, 20 Tenax and 20 Biomonitoring 5TD tubes with similar number of reconditioning cycles were identified from our pool of TD tubes and subjected to repeated (5X) reconditioning cycles and flow rate measurements.Supplemental data 6(C) show that flow rate does not largely change due to repeated reconditioning.However, Tenax TD Tube #9 showed a single low flow rate.Overall, the data suggest that the variability in TD tube flow rate present among both Tenax and Biomonitoring 5TD tubes does not seem to be related to the age, the mass of the TD tubes or empty TD tube casings, the number of reconditioning cycles, or repeated reconditioning.
Typically, transfer and concentration of volatiles from samples within breath bags onto TD tubes is performed using a pump calibrated for flow rate from a single representative blank TD tube.The flow rate, determined from the representative single tube, is used to calculate the amount of time to pass a desired volume from the bag through the TD tube via the pump.As illustrated in figure 3(A), there is a wide range (Tenax: ∆ min/max 31.1 ml min −1 , 5TD ∆ min/max 34.5 ml min −1 ) in measured flow rates among TD tubes.As all volatile loading is determined from a single representative TD tube, the illustrated variability in TD tube flow rates could induce loading inaccuracies and ultimately reporting of inaccurate results.For example, if the mean flow rate from the 5TD tubes (figure 3(A)) were used as the representative flow rate for loading (239.8 ml min −1 ) 550 mL from a bag, the 5TD tube showing the maximum flow rate (255.1 ml min −1 ) would oversample the bag by approximately 35.2 mL or 6.4% (35.2/550).Furthermore, if the lowest measured flow rate (220.68 ml min −1 ) was used as the representative rate for the same scenario, the tube showing maximum flow rate (255.1 ml min −1 ) would oversample the bag by 86.0 ml or 15.6% (86.0/550).Such inaccuracies plausibly result in variations in the data due to loading, and significantly impact resulting data.However, engineering approaches are emerging to provide simple, accurate sampling from exhaled breath bags on to TD tubes, such as that described by Chew et al [32].
The three factors that were hypothesized to impact the measured flow rates of the packed TD tubes, as shown in figure 3 and supplemental data 6, indicate there is no impact to the flow rate based on age of the TD tubes, the mass of the TD tubes or empty TD tube casings, the number of reconditioning cycles or repeated reconditioning.Therefore, it is hypothesized that the variability in flow rate could be related to how the adsorbent is 'packed' in the casing.Previous work with Tenax TA TD tubes under various extreme conditions illustrated G force had a significant impact on standard (chlorobenzene-d5 and bromofluorobenzene) peak area [33].While flow rate following G force was not tested, it is possible that the reduction in the standard was a result of lower flow rate resulting from centrifugation.Independent of the cause, the data illustrate variability in flow rate among TD tubes suggesting possible inaccuracies in sample loading are plausible.
The results provided in figure 3 and supplemental data 6 suggest variability in measured flow rates among TD tube types.Therefore, it is conceivable that flow rate variability could have an impact on TD-GC-MS results.To determine the influence differing flow rates among TD tubes has on compound peak areas, a selection of Tenax and Biomonitoring 5TD tubes that have high, middle, or low flow rates were tested from the TD tubes analyzed previously.Fifteen tubes of each flow rate and tube type were selected for a total of 45 Tenax and 45 Biomonitoring 5TD tubes.The mass, number of reconditioning cycles, and measured flow rates of each group were evaluated (supplemental data 7).The results show insignificant differences among the selected tubes, by each type, for mass (p ⩾ 0.2340) and reconditioning cycle (p ⩾ 0.6660) while a significant difference was observed for the measured flow rates (Tenax: p < 0.0001, 5TD: p < 0.0001, supplemental data 7).These data established measured flow rate is the primary source of variability of the attributes examined among the selected TD tubes making these tubes useful for further experimentation.
Utilizing 40 l bags with two separate TO-15 concentrations, 50 ppbv (high concentration) and 7 ppbv (low concentration), the low, medium, and high flow rate tubes were loaded using a representative calibrated flow rate as described previously.The TD-GC-MS results were monitored for internal standard (IS) normalized peak areas of eight selected compounds, across the entire chromatographic separation and provided in supplemental data 8.By combining the normalized peak areas from tubes representing all the flow rates (low, medium, and high) approximately 63% (20/32 measurements) of the %RSDs fall at or below 10% regardless of the concentration of sample bag (high or low, supplemental data 8).When the normalized peak areas are grouped by flow rate, similar results are observed (supplemental data 9).Supplemental data 8 show the PCA for the high (8C) and low (8D) concentration bags parsed by TD tube type and measured flow rate.A high amount of explained variation, ⩾83.8%, in the data is accounted for by PC1 and PC2.Additionally, the results demonstrate a high amount of overlap in the measured flow rates (low, medium, high) separated by TD tube type (Tenax and Biomonitoring 5TD) for both standard concentrations tested (50 ppbv supplemental data 8(C), 7 ppbv supplemental data 8(D)).Individual PCAs separated by standard concentration and TD tube type show a high amount of variation is accounted for by PC1 and PC2, ⩾76.8%, with significant overlap in the data sets suggesting high similarity among the results (supplemental data 10).Overall, the results indicate that although flow rates are statistically different among the tubes, there is a relatively small effect on the normalized peak areas of standard compounds at two distinct concentrations.

Exhaled breath measurements
While the TO-15 standard compounds are advantageous for uniformity of sample, most of the compounds contained in the standard mix are not often identified in exhaled breath.To examine the impact of TD tubes with variable measured flow rates (high, medium, and low) in a more realistic sample, a mint experiment was performed.Two compounds, tentatively identified as MMECH and methyl salicylate, were determined to be attributed to the mint from headspace analysis and were used for further monitoring in the exhaled breath (supplemental data 11).The results, shown in figure 4(A), suggest no statistically significant result (p ⩾ 0.7420) is observed between the two compounds from the mint related compounds on either TD tube type or measured flow rates (high, medium, and low) while large variability is observed among the normalized peak areas (figure 4(A)).In addition to mint related compounds, compounds historically noted as endogenous in breath, exhaled isoprene and acetone, were monitored (figure 4(B), [18]).The results show, like the mint related compounds, that there is no statistical difference among compounds, TD tubes, or TD tube flow rate.PCAs were generated from a global feature extraction of the mint data (48 features, extraction was performed on retention times less than those related to mint, <18 min).The PCAs illustrate a relatively low amount of overall variability is accounted for in PC1 and PC2 for Tenax (figure 4(C), 32.9%) and Biomonitoring 5TD (figure 4(D), 33.0%).The result is further demonstrated by the high amount of overlap in the 95% confidence intervals within the PCAs among the different tube measured flow rates.These results suggest a high amount of similarity in the datasets, where one specific principal component, i.e. variability, is spread among all the principal components and does not explain a majority of the variability in the data.Collectively, both gas standard and exhaled breath analysis with TD tubes that have high, medium, and low flow rates show similar peak areas.Therefore, while measured flow rates on TD tubes are variable, the evidence indicates a low impact to the overall results.These data suggest another uncontrolled and unknown sources of variability are plausibly impacting the data and worth further investigation.
The mint data showed large variability in the IS normalized abundances of the MMECH and methyl salicylate.It is hypothesized to be a result of the time of dissolution and/or variable oral metabolism of the mint among individuals.For instance, it was noted that the time for the mint to dissolve among the subjects ranged from 30 s to nearly 5 min.It is possible that the difference in dissolution time was related to the exhaled concentration i.e. the slower the dissolution of the mint caused a higher concentration due to the shorter off-gassing time.Regardless, the data illustrate that while variability exists in the data, both standards and endogenous and exogenous exhaled breath compounds do not show a significant difference based on the measured flow rate of the TD tube.
Overall, the results indicate that, as hypothesized, there are inconsistencies observed among TD tubes.The differences in the physical properties between TD tubes of a given type (i.e.flow rate, mass, etc) do not significantly impact the abundances of both endogenous and exogenous exhaled breath metabolites.While it is currently unknown what is contributing to the lack of in difference between TD tubes with varied flow rates, it is plausible that the conditions tested were not extreme enough to elicit a significant response.Additional experimentation surrounding the extremes of the flow range, such as using the low flow rate to load the high flow rate tubes, could plausibly induce a significant difference in response among the groups.Regardless, researchers should be aware of the potential impact of sampling techniques and the equipment utilized in exhaled breath research and evaluate their individual approaches.

Conclusion
Due to the frequently low levels of exhaled breath compounds often detected in exhaled breath analysis, investigation of sources of variation attributed to sampling media and the sampling process require constant investigation and improvement.As the exhaled breath community begins to embrace the need for guidelines for sampling and analysis of online and offline exhaled breath, consideration of the background and processes that impact sampling need to be considered.The results presented here establish the background of compounds found in two common TD adsorbent tubes, document the variation in the measured flow rate of TD tubes, and investigate the plausible sources of flow rate variation.It is hypothesized that the variation is likely due to the packing of the adsorbent within the tube.All results present herein were collected using TD tubes that were purchased from one manufacturer and it is reasonable that similar tubes purchased from another manufacturer may have different results.While flow rate variation was observed, exhaled breath data monitoring endogenous, exogenous, and global features show minimal impact of TD tube flow rate on overall abundances.

Figure 1 .
Figure 1.Comparison of background compounds identified in varying TD tube types.(A) Representative total ion chromatograms (TIC) from empty stainless steel (black), Tenax (dark gray), and biomonitoring 5TD (light gray) tubes.(B) A four-way Venn diagram illustrating the overlap in background compounds identified from each TD tube type.The data illustrate the Biomonitoring 5TD tubes have the highest abundance and number of TD tube related (background) compounds.

Figure 2 .
Figure 2.Tentatively identified background compounds compared to exhaled breath data.(A) A comparison of the peak areas of selected background compounds found in empty Tenax tubes or in 550 ml of lower airway breath loaded onto Tenax tubes.(B) A comparison of the peak areas of selected background compounds found in empty Biomonitoring 5TD tubes or in 550 ml of lower airway breath loaded onto Biomonitoring 5TD tubes.Percentages indicate the mean contribution (%) of blank areas to exhaled breath areas for statistically significant (p < 0.05) select compounds.Compounds without percentages are not statistically different (p > 0.05).The data show several compounds reported in the exhaled breath literature are statistically similar to TD tube background peak areas.Additionally, those select compounds showing a statistical difference between blanks and breath could have greater than 70% of the area contributed from the TD tube background.

Figure 3 .
Figure 3. Flow rate characteristics of Tenax and Biomonitoring 5TD tubes.(A) A scatter plot of the mean flow rates, for Tenax and Biomonitoring 5TD tubes, plotted by age of the tube.(B) The measured masses, for Tenax and Biomonitoring 5TD tubes, plotted by age of the tube.(C) A plot of the mean flow rate by measured mass for Tenax and Biomonitoring 5TD tubes.(D) A plot of the mean flow rate by number of reconditioning cycles for Tenax and Biomonitoring 5TD tubes.For all plots: dark grey indicates Tenax and light grey indicates Biomonitoring 5TD tubes with n = 200 each.The data show large variability in the flow rates among Tenax and Biomonitoring 5TD tubes.The results illustrate the variability in flow rates among TD tubes is likely not attributed to reconditioning cycles or the mass of the tubes.

Figure 4 .
Figure 4. Variability of endogenous and exogenous exhaled compounds from TD tubes shown to have variable flow rates.(A) A box plot of the internal standard (IS) normalized abundance of 1-methyl-4-(1-methylethyl) cyclohexanol (MMECH) and methyl salicylate from an exhaled breath mint utilizing tubes previously identified to have high (H) medium (M) and low (L) measured flow rates.(B) A box plot of the internal standard (IS) normalized abundance of isoprene and acetone from an exhaled breath mint experiment utilizing TD tubes previously identified to have high (H) medium (M) and (L) measured flow rates.PCA analysis of 46 features a global exhaled breath analysis following mint ingestion using TD tubes (C) Tenax, (D) Biomonitoring 5TD, determined to have high (H) medium (M) and low (L) flow measured rates.The data show both endogenous and exogenous compounds have similar results independent of the measured flow rate of the TD tubes.